When developing robust software systems, ensuring reliable and efficient testing is paramount. Yet, testing can become challenging when the System Under Test (SUT) depends on components that are unavailable, slow, or impractical to use in the testing environment. Enter Test Doubles—a practical solution to streamline testing and simulate dependent components.

What are Test Doubles? In essence, Test Doubles are placeholders or “stand-ins” that replace real components (known as Depended-On Components, or DOCs) during tests. Much like a stunt double in a movie scene, Test Doubles emulate the behavior of the real components, enabling the SUT to function seamlessly while providing better control and visibility during testing.

The implementation of Test Doubles is tailored to the specific needs of a test. Rather than perfectly mimicking the DOC, they replicate its interface and critical functionalities required by the test. By doing so, Test Doubles make “impossible” tests feasible and expedite testing cycles.

Key Variations of Test Doubles Test Doubles come in several forms, each designed to address distinct testing challenges:

1. Test Stub: Facilitates control of the SUT’s indirect inputs, enabling tests to explore paths that might not otherwise occur.
2. Test Spy: Combines Stub functionality with the ability to record and verify outputs from the SUT for later evaluation.
3. Mock Object: Focuses on output verification by setting expectations for the SUT’s interactions and validating them during the test.
4. Fake Object: Offers simplified functionality compared to the real DOC, often used when the DOC is unavailable or unsuitable for the test environment.
5. Dummy Object: Provides placeholder objects when the test or SUT does not require the DOC’s functionality.

When to Use Test Doubles Test Doubles are particularly valuable when:

• Testing requirements exceed the capabilities of the real DOC.
• Test execution is hindered by slow or inaccessible components.
• Greater control over the test environment is necessary to assess specific scenarios.

That said, it’s crucial to balance the use of Test Doubles. Excessive reliance on them may lead to “Fragile Tests” that lack robustness and diverge from production environments. Therefore, teams should complement Test Doubles with at least one test using real DOCs to ensure alignment with production configurations.

Conclusion Test Doubles are indispensable tools for efficient and effective software testing. By offering flexibility and enhancing control, they empower developers to navigate complex testing scenarios with ease. However, judicious use is key, striking the right balance ensures tests remain meaningful and closely aligned with real-world conditions.

This information comes from this article:
Test Double at XUnitPatterns.com

From the blog CS@Worcester – aRomeoDev by aromeo4f978d012d4 and used with permission of the author. All other rights reserved by the author.

One of the most significant ways of ensuring that an application is reliable and efficient before deployment is through software testing. One of the most powerful functional testing techniques that focuses on testing the boundary cases of a system is Boundary Value Analysis (BVA). Boundary Value Analysis finds potential defects that are apt to show themselves on input partition boundaries.

What is Boundary Value Analysis?

Boundary Value Analysis is a black-box testing method which tests the boundary values of valid and invalid partitions. Instead of testing all the possible values, the testers focus on minimum, maximum, and edge-case values, as these are the most error-prone. This is because defects often occur at the extremities of the input ranges rather than at any point within the range.

For example, if a system accepts values between 18 and 56, instead of testing all the values, testers would test the below-mentioned values:

Valid boundary values: 18, 19, 37, 55, 56

Invalid boundary values: 17 (below minimum) and 57 (above maximum)

By running these primary test cases, the testers can easily determine boundary-related faults without unnecessary repetition of in-between value testing.

Implementing BVA: A Real-World Example

To represent BVA through an example, let us take a system processing dates under the following constraints:

Day: 1 to 31

Month: 1 to 12

Year: 1900 to 2000

Under Single Fault Assumption, where one of the variables is tested while others are at nominal values, test cases like below can be written:

Boundary value checking for years (e.g., 1900, 1960, 2000)

Boundary value checking for days (e.g., 1, 31, invalid cases like 32)

Checking boundary values for months (i.e., 1, 12)

By limiting test cases to boundary values, we are able to have maximum test coverage with minimum test effort.

Equivalence Partitioning and BVA together

Another helpful technique is combining BVA and Equivalence Partitioning (EP) together. EP divides input data into equivalent partitions where every equivalence class is expected to behave in the same way. By using these techniques together, testers can reduce the number of test cases but still maintain complete coverage.

For instance, if a system would only accept passwords of 6 to 10 characters long, test cases can be:

0-5 characters: Not accepted

6-10 characters: Accepted

11-14 characters: Not accepted

This mix makes the testing more efficient, especially when using more than one variable.

Limitations of BVA

Although BVA is strong, it does face some limitations:

It works well when the system contains properly defined numeric input ranges.

It has no regard for functional dependencies of variables.

It may not be equally effective on free-form languages like COBOL, which has more flexible input processing.

Conclusion

Boundary Value Analysis is one very important test method that can help testers define most probable fault sites of a system. Merged with Equivalence Partitioning, it has highest test effectiveness at the maximum elimination of test case replication and minimum complete loss of test coverage. In as much as BVA isn’t a “catch-all”, yet it represents an essential technique of software provision quality and dependability.

Personal Reflection

Learning Boundary Value Analysis has helped me understand more about software testing and how it makes the software reliable. It has shown me that by focusing on boundary values, defects can be detected with higher efficiency without generating surplus test cases. It is a very practical approach to apply in real-world scenarios, such as form validation and number input testing, where boundary-related errors are likely to be found. In the future, I will include BVA in my testing approach to offer more test coverage in software projects that I undertake.

Citation

Geeks for Geeks. (n.d.). Software Testing – Boundary Value Analysis. Retrieved from https://www.geeksforgeeks.org/software-testing-boundary-value-analysis/

From the blog CS@Worcester – Maria Delia by Maria Delia and used with permission of the author. All other rights reserved by the author.

The blog I chose to write about this week details what different types of tests do and how they should be prioritized to test efficiently. The blog post starts with the writer talking about their own experience testing and how early on they tested their program only through the user interface, which quickly showed the downsides of this method. It was slow, couldn’t be run on all devices, and needed manual checks. Testing like this is called end to end testing and it is, slow, expensive, and not always the most revealing about potential problems. Instead, unit tests are much preferred as they test the very basic functionality and are quick. A middle ground between these two is instance or integration testing, which are able to cover most of the program without having to go through a limited user interface. These three types of testing create a pyramid to show that a majority of your cases should be unit tests, then instance tests being next most common, and finally end to end tests.

The distribution pyramid is based on five principles that make up the Smurf mnemonic. The first is speed, that has you prioritize many quick tests that will catch problems sooner. Next is maintainability that gives importance to tests that will scale well or not be subject to many dependencies. Utilization is important when keeping in mind the cost of running your code repetitiously and minimizing the resources used. Reliability states that you should always have tests that will only return an error when something is actually wrong and using these as indicators of crucial issues that need to be addressed. Lastly, fidelity is testing that recreates the users experience from start to finish, or end to end. Each type of test in the pyramid is distributed across these five factors in variable amounts, each showing their use.

I chose this blog because I wanted to learn more about writing test cases for a complete program in a work environment. I thought that this blog did well in that respect and helped in so far as providing an outline to begin with when starting to write test code. One addition that could have improved the post would be some examples, but they are easily accessible elsewhere. This will be a helpful resource and reference to use in the future when I am put in the position where I would need to start writing test from scratch, as well as being something to keep in mind when looking at prewritten tests to compare.

Test Pyramid Google Testing Blog: SMURF: Beyond the Test Pyramid

From the blog CS@Worcester – Computer Science Blog by dzona1 and used with permission of the author. All other rights reserved by the author.

For a lot of the projects I’ve worked on, we didn’t really do testing in the traditional way of actually writing tests. Each method or function we created, we would just use debugging in the actual code itself to figure out which parts of the method is working correctly, which fail safes are activating, and which errors are throwing. There didn’t seem like a whole reason for writing tests using JUnit or something similar when all code went to a dev environment first, and then put to production.

Just recently I started using JUnit a little bit just to find out how useful it could be when practicing coding certain problems. It definitely had some perks to it such as without compiling and building the entire project each time, I could just run the test to check if that certain code piece will function properly or not. But for me, that isn’t really enough to warrant using it all that much, if you have access to a dev environment (which you should), even though it isn’t the “proper” way to do things, I believe that writing debug in code instead of tests, is way more efficient.

If you look at it from a different perspective though, writing tests can also be worthwhile if you’re more into the automation part of developing. Since tests will be ran automatically when compiling your projects, you wouldn’t need to keep adding in more debug statements manually. This could increase efficiency more than what I previously mentioned, but you may not get as much information as needed from the test. If one of your methods isn’t working properly but the test keeps returning an exception or error and you don’t understand the cause of why it’s happening, the previous approach of writing debug statements will be more efficient and in turn, will help you understand where you went wrong.

All in all, I personally believe that both unit testing / testing in general, and writing debugs have their own places when writing software. Obviously I don’t know too much about software testing yet, but the more I learn, there is definitely potential that my opinion on it would change. As of now, writing debugs if there is an issue, or even if there isn’t, seems to help me a lot when understanding each part of what I’m writing in a deeper level and I look forward to learning more about testing to where it could also help me reach that point.

From the blog CS@Worcester – CS Blog by Mike and used with permission of the author. All other rights reserved by the author.

This week in class we learned about path testing, which is a white box method that examines code to find all possible paths. Path testing uses control flow graphs in order to illustrate the different paths that could be executed in a program. In the graph, the nodes represent the lines of code, and the edges represent the order in which the code is executed. Path testing appealed  to me as a testing method because it gives visual representations of how the source code should execute given different inputs. I took a deeper dive into path testing after this week’s classes and found this blog that gave me a deeper understanding of path testing.

Steps

When you have decided that you want to perform path testing, you must create a control flow graph that matches up with the source code. For example, the split of direction between nodes should represent if-else statements and for while loops, the nodes towards the end of the program that have an edge pointed at an earlier node. 

Secondly, pick out a baseline path for the program. This is the path you define to be the original path of your program. After the baseline is created, continue generating paths representing all possible outcomes in the execution. 

How many Test Cases?

For a lengthy source code, the possible outcomes could seem endless and could therefore end up being a difficult, time-consuming task to do manually. Luckily, there is an equation that determines how many test cases a program will need with path testing.

C = E – N + 2P

Where C stands for cyclomatic complexity. The cyclomatic complexity is equivalent to the number of linearly independent paths, which in turn equals the number of required test cases. E represents the number of edges, N is the number of nodes, and P is the number of connected components. Note that for a single program or source of code, P = 1 always.

Benefits

Path testing reveals outcomes that otherwise may not have been known without examining the code. As stated before, it can be difficult for a tester to know all the possible outcomes in a class. Path testing provides a solution to that by using control flow charts, where the tester can examine the different paths. Path testing also ensures branch coverage. Developers don’t need to merge code with an existing repository because the developers can test in their own branch. Unnecessary and overlapping tests are another thing developers don’t have to worry about.

Drawbacks

Path testing can also be time consuming. Quicker testing methods do exist and take less time off further developing projects. Also in many cases, path testing may be unnecessary. Path testing is used often by many DevOps setups that require a certain amount of unit coverage before deploying to the next environment. Outside of this, it may be considered inefficient compared to another testing method.

Blog: https://blog.testlodge.com/basis-path-testing/

From the blog Blog del William by William Cordor and used with permission of the author. All other rights reserved by the author.

I chose this blog because I was interested in learning how to optimize the selection of test cases without sacrificing thorough coverage. During my search, I came across an article on TestGrid.io on equivalence partitioning testing, which was fantastic in removing redundancy from testing. As I develop my programming and software testing skills, I have found this method especially useful in making the testing process simpler.

Equivalence partitioning is a testing technique applied to partition input data into partitions or groups based on the assumption that values in each partition will behave similarly. Instead of testing each possible input, testers choose a sample value from each partition and hope the entire group will result in the same output. This reduces the number of test cases but still provides sufficient coverage.

For example, if a program is able to accept input values ranging from 1 to 100, equivalence partitioning allows the testers to categorize them into two sets: valid values (1-100) and invalid values (less than 1 or more than 100). Rather than testing every number in the valid set, a tester would choose sentinel values like 1, 50, and 100. Similarly, they would test the invalid range with 0 and 101. This is time-efficient but identifies errors simultaneously.

I employed the TestGrid.io article because it explains equivalence partitioning in an understandable and systematic manner. Much other testing material is too complex or ambiguous for newcomers, but this article simplifies it and incorporates real-world examples. This allowed it to be simple not only to understand the theory, but also to apply the method to real-life situations.The article also discusses the advantages of equivalence partitioning, including reducing redundant test cases, being more efficient, and offering complete coverage. As an individual interested in improving my testing methods, I found this useful because it corresponds with my goal of producing better, more efficient test cases without redundant repetition.

Equivalence partitioning testing is a sound approach to maximizing test case selection. It enables the tester to focus on representative cases rather than testing all possible inputs, which is time- and effort-efficient. The TestGrid.io article provided a clear understanding of how to implement this method and why it is significant. For me, learning effective test methods like equivalence partitioning will make me more efficient in my coding, debugging, and software developing abilities, prepared for internships, projects, and software engineering positions.

Blog: https://testgrid.io/blog/equivalence-partitioning-testing/

From the blog CS@Worcester – Matchaman10 by tam nguyen and used with permission of the author. All other rights reserved by the author.

This week, our class learned about path testing. It is a white box testing method (meaning we do not actually run the source code) that tests to make sure every specification is met and helps to create an outline for writing test cases. Each line of code is represented by a node and each progression point the compiler follows is represented by a directed edge. Test cases are written for each possible path the compiler can take.

For example, if we are testing the code block below, a path testing diagram can be created: also shown below. Each node represents a line of code and each directed edge represents where the compiler goes next, depending on the conditions. Test cases are written for each condition: running through the while loop and leaving when value is more than 5 or bypassing it if value starts as more than 5.

I wanted to learn more about path testing, so I did some research and found a blog that mentioned cyclomatic complexity. Cyclomatic complexity is a number that classifies how complex the code is based on how many nodes, edges, and conditionals you have. This number will relate to how many tests you need to run, but is not always the same number. The cyclomatic complexity of the example above would be (5-5)+2(1) = 2.

Cyclomatic Complexity = Edges – Nodes + 2(Number of Connected Components)

The blog also explores the advantages and disadvantages of path based testing. Some advantages are performing thorough testing of all paths, testing for errors in loops and branching points, and ensuring any potential bugs in pathing are found. Some disadvantages are failing to test input conditions or runtime compilations, a lot of tests need to be written to test every edge and path, and exponential growth in the number of test cases when code is more complex.

Another exercise we did in class was condensing nodes that do not branch. In the example above, node 2 and 3 can be condensed into one node. This is because there is no alternative path that can be taken between the nodes. If line 2 is run, line 3 is always run right after, no matter what number value is. Condensing nodes would be helpful in slightly more complex programs to make the diagram more readable. Though if you are working with a program with a couple hundred lines, this seems negligible.

When I am writing tests for a program in the future, I would probably choose a more time conscious method. Cyclomatic complexity is a good and useful metric to have, but basing test cases off of the path testing diagram does not seem practical for complex codes and tight time constraints.

Blog post referenced: https://www.testbytes.net/blog/path-coverage-testing/

From the blog CS@Worcester – ALIDA NORDQUIST by alidanordquist and used with permission of the author. All other rights reserved by the author.